Customer Stories

posted on 18 April 2008 08:33

BT has undergone a wholesale and massive virtualisation exercise in its datacentres and their servers and storage with astounding results:

1. Eight datacentres were consolidated to five.
2. 660 racks were elimated leaving just 40.
3. Of 3,100 Wintel servers 2,966 were eliminated with a 15:1 consolidation ratio.
4. Full backup run time shrank from 96 hours to 30 minutes.
5. Over 8,000 network ports were eliminated, leaving 840.
6. A 2.1 megawatt power draw was lowered to less than a quarter of a megawatt.
7. Server deployment, which used to take six weeks, reduced to one day.
8. It disposed of 250 tonnes of unwanted equipment.

The driving causes were space, cooling capacity and power limitations with BT needing 0.7 percent of the total UK power generating capacity in 2006 and paying heavily for that. It was too expensive to build more data centres and IT was failing to meet its service level agrerements (SLA). The answer was to do more with less.

BT wanted to improve access to stored information, lower datacentre operating costs, increase server and storage utilisation and use a capacity-on-demand scheme to answer dynamic business needs.

The two main datacentre reduction technologies were VMware server virtualisation and NetApp-based storage area networks (SAN).

Michael Crader was the Windows consolidation head at BT and said it was a 2-year project. Most servers had their own direct-attached storage and both server and storage utilisation was poor. VMware sorted out the server situation and a SAN was needed to consolidate storage and make the virtualised servers work.

BT synchronised the virtualisation of the servers and their storage. It standardised on NetApp for its storage and did so partly because it enabled thin provisioning and thus increased storage utilisation, which rose to 70 percent. The NetApp products also supported the three main storage protocols: Fibre Channel; iSCSI; and NFS, giving BT connection flexibility. The Fibre Channel usage prevented storage I/O bottlenecks.

Because of NetApp's snapshot technology BT was able to get rid of tape backups. Disk to disk backup is used instead, with a large number of virtual snaps are used. It was also able to replicate data across its wide area network and thus provide business continuity and disaster recovery. File restoration time shrank from hours and even days down to seconds. RAID 6 technology provided the disk failure protection needed.

It saved massive amounts on its power bills, drastically reduced maintenance contracts, redeployed or fired unwanted technical staff and generally improved IT department SLA performance. Financial payback was achieved by Christmas, 2007.

It wasn't a simple project by any means and one way it succeeded was by only using VMware-certified products. There was a through re-examination of the datacentres' physical infrastructure: servers; storage; and storage networking. There was also a thorough look at and overhaul of the storage provisioning, data protection, compliance areas and more.

BT succeeded remarkably in doing far more productive IT work with far less physical IT kit through using server and storage virtualisation in concert. It progressed its green agenda as well. Where BT has gone other organisations lacking combined and virtualised server and storage infrastructures should follow.

^{[Chris Mellor. Thanks to Wikibon.]}